Quantitative analysis of neuronal morphology via machine learning

Encadrants : Suvadip Mukherjee
Disponible : NON
Spécialité : IMA
Nombre d'étudiants : 1
Description :

Deciphering the relationship between the morphology (structure) and the function of neurons remain an unsolved open problem in modern neuroscience. Recently, there have been elaborate studies to create digital reconstruction of neurons from digital microscopy using sophisticated tools from image analysis and machine learning [1]. The current focus is to analyze these digital reconstructions at scale, to further our understanding about the interplay between the geometry of the neurons and their functional characteristics.

Pré-requis : This project does not require the student to be familiar with either biology or neuroscience. However, preliminary understanding of pattern recognition/signal processing techniques, and a strong background in applied mathematics is desirable. Additionally, proficiency in at least one high level programming language such as Python/Java/C++ will be necessary.
Travail demandé : Project plan:This projectaims at developing machine learning techniques to characterize and classify the digitally reconstructed neurons for automated shape-based classification and retrieval [2]. Similar to the image-based retrieval feature in popular internet search engines, wewish to develop a supervised learning algorithm to retrieve neurons from a large atlas of neurons which is available athttp://neuromorpho.org/[3].References1. Peng, Hanchuan, et al. "BigNeuron: large-scale 3D neuron reconstruction from optical microscopy images." Neuron 87.2 (2015): 252-256.2. Basu, Saurav, Barry Condron, and Scott T. Acton. "Path2Path: Hierarchical path-based analysis for neuron matching." 2011 IEEE International Symposium on Biomedical Imaging: From Nano to Macro. IEEE, 2011.3. Kervrann, Charles, et al. "A guided tour of selected image processing and analysis methods for fluorescence and electron microscopy." IEEE Journal of Selected Topics in Signal Processing 10.1 (2015): 6-30.